Abstract
The implementation of statistical postprocessing of ensemble forecasts is increasingly developed among national weather services. The so-called Ensemble Model Output Statistics (EMOS) method, which consists of generating a given distribution whose parameters depend on the raw ensemble, leads to significant improvements in forecast performance for a low computational cost, and so is particularly appealing for reduced performance computing architectures. However, the choice of a parametric distribution has to be sufficiently consistent so as not to lose information on predictability such as multimodalities or asymmetries. Different distributions are applied to the postprocessing of the European Centre for Medium-range Weather Forecast (ECMWF) ensemble forecast of surface temperature. More precisely, a mixture of Gaussian and skewed normal distributions are tried from 3- up to 360-h lead time forecasts, with different estimation methods. For this work, analytical formulas of the continuous ranked probability score have been derived and appropriate link functions are used to prevent overfitting. The mixture models outperform single parametric distributions, especially for the longest lead times. This statement is valid judging both overall performance and tolerance to misspecification.
Highlights
In numerical weather prediction (NWP), ensemble forecasts aim to capture the expected uncertainty associated with a weather forecast [1]
This study focuses on local implementation of Ensemble Model Output Statistics (EMOS)
Probabilistic calibration is commonly evaluated with probability integral transform (PIT) or rank histograms (RH) [49,50,51,52]
Summary
In numerical weather prediction (NWP), ensemble forecasts aim to capture the expected uncertainty associated with a weather forecast [1]. This paper deals with the widely used technique called nonhomogeneous regression or ensemble model output statistics (EMOS) [15] This type of distribution regression approach has been extended to numerous variables’ responses and distributions such as Gaussian distributions [15,16], truncated normal [17], log-normal [18], gamma [19], or generalized extreme values [20,21]. The seminal work of Gneiting et al [15] considered Gaussian distributions for the EMOS postprocessing of ensemble forecasts of surface temperature. Mixture and single parametric Gaussian and skew normal distributions are compared on the ECMWF ensemble forecast of surface temperature, using different estimation methods.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
